[
https://issues.apache.org/jira/browse/HIVE-19357?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16458250#comment-16458250
]
Matt McCline commented on HIVE-19357:
-------------------------------------
Ya, the VectorUDFAdaptor class catches any Exception from evaluating the
generic UDF and suppresses it to NULL. So, the question becomes – is this a
good policy? I think no.
{code:java}
// call function
Object result;
try {
result = genericUDF.evaluate(deferredChildren);
} catch (HiveException e) {
/* For UDFs that expect primitive types (like int instead of Integer or
IntWritable),
* this will catch the the exception that happens if they are passed a
NULL value.
* Then the default NULL handling logic will apply, and the result will
be NULL.
*/
result = null;
}{code}
> vectorization breaks assert functionality
> -----------------------------------------
>
> Key: HIVE-19357
> URL: https://issues.apache.org/jira/browse/HIVE-19357
> Project: Hive
> Issue Type: Bug
> Components: Vectorization
> Reporter: Zoltan Haindrich
> Priority: Major
>
> This could be limited to assert exceptions; but might interfere with other
> exceptions...discovered while "fixing" testreopt after HIVE-19269
> {code}
> create table tu(id_uv int,id_uw int,u int);
> create table tv(id_uv int,v int);
> create table tw(id_uw int,w int);
> insert into tu values
> (10,10,10),(1,1,1),(2,2,2),(3,3,3),(4,4,4),(5,5,5),(6,6,6);
> insert into tv values (10,10),(1,1),(2,2),(3,3);
> insert into tw values
> (10,10),(1,1),(2,2),(3,3),(4,4),(5,5),(6,6),(7,7),(8,8),(9,9);
> set zzz=0;
> set hive.vectorized.execution.enabled=false;
> select assert_true(${hiveconf:zzz}>sum(1)) from tu join tv on
> (tu.id_uv=tv.id_uv) where u<10 and v>1;
> -- fails as expected
> set hive.vectorized.execution.enabled=true;
> select assert_true(${hiveconf:zzz}>sum(1)) from tu join tv on
> (tu.id_uv=tv.id_uv) where u<10 and v>1;
> -- there is a result set
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
